1 Statistics 101-106 Lecture 12 (1 Dec 98) David Pollard Lecture 12 Each year the U.S. News and World Report provides rankings of US colleges (and universities), based on data provided by the colleges themselves. The web page at http://www.usnews.com/usnews/edu/college/corank.htm documents (see Rankings information) the methods that US News uses. The data for the 1994 rankings (downloaded from the web site for the Journal of Statistics Education: gopher://jse.stat.ncsu.edu:70/11/jse/) are linked to the datasets page at Stat100++ web site. The data consist of the following variables: FICE = FICE (Federal ID number) name = College Name (abbreviated) State = State type = Public/private indicator (public=1, private=2) SATm = Average Math SAT score SATv = Average Verbal SAT score SAT = Average Combined SAT score ACT = Average ACT score SATm25 = First quartile - Math SAT SATm75 = Third quartile - Math SAT SATv25 = First quartile - Verbal SAT SATv75 = Third quartile - Verbal SAT ACT25 = First quartile - ACT ACT75 = Third quartile - ACT new.applied = Number of applications received new.accepted = Number of applicants accepted new.enrolled = Number of new students enrolled new.top10HS = Pct. new students from top 10% of H.S. class new.top25HS = Pct. new students from top 25% of H.S. class undergrads = Number of fulltime undergraduates parttime = Number of part-time undergraduates tuition.in = In-state tuition tuition = Out-of-state tuition room+board = Room and board costs room = Room costs board = Board costs cost.fees = Additional fees cost.books = Estimated book costs cost.pers = Estimated personal spending fac.phd = Pct. of faculty with Ph.D.’s fac.term = Pct. of faculty with terminal degree sf.ratio = Student/faculty ratio donate = Pct.alumni who donate exp.stud = Instructional expenditure per student graduate = Graduation rate College = College name (in full) A star (*) in the original data indicates a missing value. The software that I use writes NA for missing values. I extracted the subset of the data , consisting of the 1133 colleges for which there is AAUP salary data available. The U.S. News derives ranks by taking weighted combinations of scores derived from such data. For example, ‘Student selectivity’ contributes 15% to the ranking score, with average SAT/ACT scores contributing 40% towards the ‘Student selectivity’. Missing data complicate comparisons of SAT/ACT scores across colleges. For some colleges we have only first and third SAT quartiles; for others we have only average SAT scores; for others we have only ACT scores or quartiles. How should we combine these pieces of partial information into a single SAT measure?
15
Embed
Lecture 12 - Yale University1 Statistics 101-106 Lecture 12 (1 Dec 98) David Pollard Lecture 12 Each year the U.S. News and World Report provides rankings of US colleges (and universities),
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
1
Statistics 101-106 Lecture 12 (1 Dec 98) David Pollard
Lecture 12
Each year the U.S. News and World Report provides rankings of US colleges (and universities), based ondata provided by the colleges themselves. The web page at http://www.usnews.com/usnews/edu/college/corank.htm
documents (see Rankings information) the methods that US News uses.
The data for the 1994 rankings (downloaded from the web site for the Journal of Statistics Education:gopher://jse.stat.ncsu.edu:70/11/jse/) are linked to the datasets page at Stat100++ web site. The data consist ofthe following variables:
FICE = FICE (Federal ID number)name = College Name (abbreviated)State = Statetype = Public/private indicator (public=1, private=2)SATm = Average Math SAT scoreSATv = Average Verbal SAT scoreSAT = Average Combined SAT scoreACT = Average ACT scoreSATm25 = First quartile - Math SATSATm75 = Third quartile - Math SATSATv25 = First quartile - Verbal SATSATv75 = Third quartile - Verbal SATACT25 = First quartile - ACTACT75 = Third quartile - ACTnew.applied = Number of applications receivednew.accepted = Number of applicants acceptednew.enrolled = Number of new students enrollednew.top10HS = Pct. new students from top 10% of H.S. classnew.top25HS = Pct. new students from top 25% of H.S. classundergrads = Number of fulltime undergraduatesparttime = Number of part-time undergraduatestuition.in = In-state tuitiontuition = Out-of-state tuitionroom+board = Room and board costsroom = Room costsboard = Board costscost.fees = Additional feescost.books = Estimated book costscost.pers = Estimated personal spendingfac.phd = Pct. of faculty with Ph.D.’sfac.term = Pct. of faculty with terminal degreesf.ratio = Student/faculty ratiodonate = Pct.alumni who donateexp.stud = Instructional expenditure per studentgraduate = Graduation rateCollege = College name (in full)
A star (*) in the original data indicates a missing value. The software that I use writes NA for missingvalues.
I extracted the subset of the data , consisting of the 1133 colleges for which there is AAUP salary dataavailable.
The U.S. News derives ranks by taking weighted combinations of scores derived from such data. Forexample, ‘Student selectivity’ contributes 15% to the ranking score, with average SAT/ACT scorescontributing 40% towards the ‘Student selectivity’.
Missing data complicate comparisons of SAT/ACT scores across colleges. For some colleges we haveonly first and third SAT quartiles; for others we have only average SAT scores; for others we have onlyACT scores or quartiles. How should we combine these pieces of partial information into a single SATmeasure?
2
Statistics 101-106 Lecture 12 (1 Dec 98) David Pollard
Statistics 101-106 Lecture 12 (1 Dec 98) David Pollard
For all except 2 colleges with SAT present, both SATm and SATv are also present. For only 1 college withboth SATm and SATv present, is SAT missing. In most cases SAT = SATm +SATv. No college has onlytwo or three of the four SAT quartiles, or only one of the ACT quartiles.